Some things are built to be useful. Others help us remember why we build.

01Policy-Relevant Science & Technology

Munk Debate: Does AI Pose an Existential Threat?

Yoshua Bengio and Max Tegmark vs. Yann LeCun and Melanie Mitchell

Probably the best single introduction to the central intellectual divide in AI. Bengio and Tegmark argue that advanced AI could become uncontrollable, while LeCun and Mitchell argue that the danger is exaggerated and based on mistaken assumptions about intelligence.

Are AI extinction warnings responsible foresight, or technologically confused fearmongering?

02Yuval Noah Harari

AI Has Hacked the Code of Human Civilization

Yuval Noah Harari

Harari’s strongest argument is broader than “AI will take jobs.” He says human civilization runs on language, stories, laws, religions, currencies, and institutions, and AI is the first technology capable of independently producing the material from which those systems are built.

What happens when humans no longer control the production of culture and meaning?

03Imagination in Action

Why LLMs Will Not Lead to AGI

Yann LeCun

This is the antidote to almost every breathless prediction about scaling language models. LeCun argues that LLMs lack genuine world understanding, persistent memory, planning, and the ability to learn through interaction with physical reality.

This one is particularly relevant to my work on agentic organizations because it challenges whether today’s agents possess meaningful agency or merely perform convincing linguistic simulations.

Is the current agentic-AI movement being built on a technological architecture that has already reached its conceptual limits?

04The Royal Institution

Geoffrey Hinton: Will AI Outsmart Human Intelligence?

Geoffrey Hinton

Hinton helped create the deep-learning revolution and now believes digital intelligence may become superior to biological intelligence. His argument about knowledge-sharing is especially interesting: one AI model can transfer what it learns to countless copies almost instantly, while humans communicate slowly and imperfectly.

Is human intelligence merely an early, inefficient form of intelligence that digital systems will supersede?

05Lex Fridman

Dario Amodei on Claude, AGI and the Future of AI

Dario Amodei, Amanda Askell and Chris Olah

A deep conversation with Anthropic’s leadership about powerful AI, model character, interpretability, alignment and the possibility of building systems that exceed expert human performance.

Amodei’s vision sits in an unusual position: extremely optimistic about AI’s potential to transform science and medicine, while simultaneously warning about catastrophic risk and concentrated power.

Can the companies racing to build transformative AI also be trusted to define what “safe” AI means?

06TED

The Catastrophic Risks of AI, and a Safer Path

Yoshua Bengio

Bengio argues that the danger increases when AI systems become agentic: pursuing goals, planning, deceiving, preserving themselves and acting across environments. His proposed response is to develop non-agentic “scientist AI” systems rather than autonomous actors.

This directly intersects with the agentic-firm thesis. It raises the possibility that the most commercially attractive AI architecture may also be the most structurally dangerous.

07Center for Humane Technology

The AI Dilemma

Tristan Harris and Aza Raskin

A highly influential presentation arguing that society is deploying AI without understanding its emergent capabilities. Some experts consider it essential public education; others criticize it for selectively presenting dramatic examples and amplifying fear.

Are AI companies repeating the social-media catastrophe at a vastly larger scale?

08Doom Debates

Max Tegmark vs. Dean Ball: Should We Ban Superintelligence?

Max Tegmark and Dean Ball

This debate moves beyond generic AI safety into an explicit political question: should governments prohibit the development of superintelligent AI?

Tegmark sees prevention as rational civilizational self-preservation. Ball argues that prohibition may be technically unrealistic, geopolitically destabilizing and potentially authoritarian.

09The Curve

Debate on AI Takeover and Existential Risk

Sayash Kapoor and Daniel Kokotajlo

A more analytical version of the AI-doom debate. Kapoor represents the skeptical “AI Snake Oil” perspective, while Kokotajlo argues that rapidly advancing AI could produce takeover-level risks.

Are catastrophic AI forecasts legitimate scenario planning, or stories built on chains of fragile assumptions?

10MIT Sloan Management Review

AI Is Not Improving Productivity

Daron Acemoglu

Acemoglu challenges the assumption that more AI automatically produces prosperity. His argument is that the direction of technology is shaped by incentives. Companies may use AI to remove workers, monitor them or reduce labour costs instead of expanding human capability.

What if the current AI boom generates enormous corporate valuations without producing broad economic progress?

11Project Syndicate

Daron Acemoglu and Michael Sandel on AI, Power and Inequality

Daron Acemoglu and Michael Sandel

This discussion moves from productivity into moral and political philosophy. It examines who controls technology, who receives its gains and why “innovation” should not automatically be considered social progress.

Should society accept every economically valuable use of AI simply because it is technologically possible?

12Hard Fork

Dario Amodei: Hopes and Fears for the Future of AI

Dario Amodei

Amodei describes a near future in which AI could accelerate biology, medicine, economic development and scientific discovery, potentially compressing decades of progress into a few years. At the same time, he acknowledges severe disruption, geopolitical competition and misuse risks.

What happens to society when intelligence becomes abundant, but ownership of that intelligence remains highly concentrated?

13The Economist

AI and Our Future

Yuval Noah Harari and Mustafa Suleyman

A revealing exchange between a historian warning about civilizational disruption and an AI entrepreneur responsible for building the technology. Suleyman frames containment as a practical governance challenge; Harari questions whether institutions can adapt quickly enough.

Can AI be contained through governance, or does competitive pressure make meaningful containment impossible?